Take f(x)=x2 and nudge the input by 0.1 at a few places. From 1 to 1.1 the output goes from 1 to 1.21, a change of 0.21, so the ratio ΔxΔf is 2.1. From 5 to 5.1 it goes from 25 to 26.01, a change of 1.01, ratio 10.1. The same small nudge moves the output almost five times as much at x=5. Sensitivity is that ratio: output change per unit of input change.
On a graph, sensitivity is steepness. A straight line has the same steepness everywhere, so g(x)=3x is equally sensitive at every input: any nudge of 0.1 moves the output 0.3. Curves change steepness, so their sensitivity is local, and you must say where you measured it. As Δx shrinks, the ratio settles on a single number, the derivative f′(x), the exact sensitivity at that point.
A large output is not the same as a sensitive one. The function h(x)=1000 has huge outputs and zero sensitivity, since no nudge changes it. Meanwhile p(x)=50x near x=0 has outputs close to 0, yet every nudge of 0.1 moves it by 5. What matters is how far the output moves, not how big it is. Report the size of your nudge as well: on a curve, a ratio measured over Δx=0.1 and one over Δx=2 can differ a lot.
Think of it like this. A coarse knob and a touchy one: on the touchy knob a tiny twist makes a big jump in loudness, while on the coarse one you can turn a long way and hear little change. A curved function is a knob whose touchiness changes as you turn it, gentle at one end and sharp at the other.
Core idea. Sensitivity measures how much the output moves for a tiny change in the input — steep means very sensitive.